More The TWIML AI Podcast episodes
Published 29 Jan 2026
Duration: 3981
Language models can struggle to generate diverse responses, despite adjustments, and improving their performance requires addressing data quality and training techniques.
Today, we're joined by Yejin Choi, professor and senior fellow at Stanford University in the Computer Science Department and the Institute for Human-C...
The podcast examines how both small and large language models tend to generate similar, homogeneous outputs in response to open-ended prompts, even when temperature settings are modified to encourage diversity. It investigates research efforts aimed at improving small language models (SLMs) by refining their reasoning abilities through methods like better data curation, synthetic data generation, and hybrid model designs. The discussion points out the limitations of relying on internet-based training data and stresses the value of high-quality, specialized content created by humans to train SLMs more effectively.
The podcast also reviews techniques such as imitation learning, reinforcement learning with verification, and data filtering to enhance model performance and output diversity. Additionally, it raises broader concerns about the effects of AI on human creativity and thought, as well as the potential for AI to contribute to the homogenization of online content. Emphasis is placed on the importance of making AI more accessible, exploring diverse alignment approaches, and developing systems that are more data-efficient and ethically responsible.
16 Sept 2026 From Voice Agents to AI Avatars with Alexander Smola
"Voice AI advances toward human-like avatars but faces challenges like latency, emotional authenticity, and hardware limitations, with future systems balancing speed, multimodal training, and ethical considerations."
9 Sept 2026 Do AI Tokenomics Matter More Than Model Benchmarks? with Chris Potts
"Explores AI's economic challenges, diminishing returns of token-based models, linguistic foundations, inefficiencies in current architectures, open-source contributions, and the need for high-risk research to drive innovation."
25 Aug 2026 Why the Next AI Breakthrough May Come from Physics with Max Welling
"AI accelerates scientific research in molecular dynamics and material science through equivariant neural networks, machine learning force fields, and digital twin simulations, revolutionizing fields like carbon capture, semiconductors, and energy storage while integrating physics and machine learning for broader scientific advancements."
27 Jul 2026 Why Models Are AIs Next Training Dataset with Damian Borth
"Explores weights-based learning, treating neural network weights as input data to train new models, improving efficiency, addressing data scarcity, and enabling tasks like model compression and performance prediction, with future directions in scaling, privacy, and cross-domain knowledge transfer."
8 Jul 2026 How AI Learns to Smell with Alex Wiltschko
Digitizing scent using AI involves converting molecules into digital data, creating standardized scent representations, and reproducing odors, addressing biological complexities, leveraging graph neural networks, and exploring applications in fragrance, diagnostics, and emotion while highlighting technical and ethical challenges.